What a system’s relaxation modes can—and cannot—reveal about equilibrium

Reading equilibrium from relaxation

A disturbed system can settle through several patterns of motion, each with a characteristic decay rate and sometimes an oscillation frequency. These modes form its spectrum. In a theoretical tutorial, independent researcher Alvan Hunt asks how much that spectrum reveals about detailed balance: the equilibrium condition that each transition between states is balanced by its reverse.

The answer has an important asymmetry. For a specified class of random-transition models, some spectral features rule out detailed balance. Their absence generally does not establish it. Nor does a topological classification, by itself, tell us how much entropy the system produces.

The preprint combines mathematical arguments and worked models rather than reporting new experiments. It also rederives established results to explain connections across otherwise different systems.

What closed loops have in common

The organizing idea is holonomy: what changes when something is transported around a closed loop. A wave state can acquire a phase. A circuit through a defective crystal can fail to close by a displacement called the Burgers vector. In a network of random transitions, the cycle affinity is the natural logarithm of the product of forward-to-backward rate ratios around a cycle. It measures the cycle’s directional bias. At detailed balance, that product equals one, so its logarithm vanishes.

These quantities share mathematical structures, not necessarily microscopic causes. Hunt explicitly warns against treating similar equations as evidence that fluids, crystals and quantum systems share one underlying physical mechanism.

For random-transition networks, the loop connection has a direct thermodynamic meaning. Under the paper’s assumptions, detailed balance holds exactly when every cycle affinity vanishes. Nonzero stationary circulation then contributes to entropy production, a measure of irreversibility.

A spectral warning is not a dissipation meter

The main theorem concerns finite, connected Markov networks with a unique stationary distribution, transitions available in both directions, and states that are unchanged by time reversal. A Markov model assumes that the present state determines transition probabilities without additional memory.

Under these conditions, detailed balance requires the generator—the matrix governing probability evolution—to have real eigenvalues and a complete set of independent eigenvectors. A non-real eigenvalue or an exceptional point, where eigenvalues and eigenvectors merge, therefore rules out detailed balance. In translation-invariant rings satisfying additional assumptions, a nonzero spectral winding number does so too. This integer counts how the spectrum loops around a reference point in the complex plane.

The reverse inference fails. Hunt gives a three-state network with distinct, real eigenvalues of zero, minus four and minus five in the model’s time units, yet with a circulating stationary current. A spectrum that passes this basic equilibrium check can still conceal irreversible dynamics.

A biased hopping ring makes a second limitation explicit. If clockwise and counterclockwise rates are p and q, its stationary entropy production rate is Boltzmann’s constant multiplied by (p − q) ln(p/q). For p > q > 0, the spectral curve is an ellipse: its winding number is minus one about any reference point inside the ellipse, but zero about a point outside. Choosing the ellipse’s center, −(p + q), as the reference point for each member of the family keeps the winding at minus one while entropy production can range over every positive value as the rates vary. The integer detects an obstruction to equilibrium, not its thermodynamic price.

Quantitative bounds need more information. The tutorial invokes the thermodynamic uncertainty relation, which links entropy production to a current’s mean rate and fluctuations. Unlike a winding number alone, those measurements provide a time scale.

Exceptional points need physical context

An exceptional point is not a universal certificate of nonequilibrium. Momentum reverses sign when time runs backward; position does not. Accounting for this distinction allows an ordinary thermally equilibrated oscillator to reach an exceptional point at critical damping, where its two relaxation eigenvalues merge.

The paper contrasts that example with an overdamped odd-elastic model, whose displacement variables are unchanged under time reversal. Its unusual elastic coupling permits work over a closed strain cycle. With the specified thermal noise, this model violates detailed balance whenever that coupling is nonzero—not only at its exceptional point.

The quantum discussion likewise requires care about definitions. A two-site tunnelling model with dephasing has an exceptional point while satisfying the particular quantum detailed-balance definition used in the paper, which permits a Hamiltonian contribution. Its stationary state is maximally mixed, corresponding to infinite temperature; the example does not establish the same claim for a finite-temperature bath.

Proposed tests, bounded conclusions

Suggested tests include tracking a particle hopping around a ring and measuring a trapped particle as it crosses critical damping. These are proposals, not new experimental confirmations.

For the stable stochastic models examined, stationary entropy production remains smooth through exceptional points even though the decomposition into modes becomes singular. The practical lesson is to measure currents, fluctuations, noise and time-reversal properties alongside spectra. Spectral structure can expose incompatibility with equilibrium, but it is not a standalone thermodynamic measurement.